maggy DocumentationRelease 0.4.0

Logical Clocks AB

Apr 19, 2020

Contents:

1 Quick Start 3

2 MNIST Example 5

3 Documentation 73.1 Maggy User API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73.2 Maggy Developer API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.3 Release 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.4 License . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 Indices and tables 15

Python Module Index 17

Index 19

i

ii

maggy Documentation, Release 0.4.0

Maggy is a framework for efficient asynchronous optimization of expensive black-box functions on top of ApacheSpark. Compared to existing frameworks, maggy is not bound to stage based optimization algorithms and therefore itis able to make extensive use of early stopping in order to achieve efficient resource utilization.

For a video describing Maggy, see this talk at the Spark/AI Summit.

Right now, maggy supports asynchronous hyperparameter tuning of machine learning and deep learning models, andablation studies on neural network layers as well as input features.

Moreover, it provides a developer API that allows advanced usage by implementing custom optimization algorithmsand early stopping criteria.

To accomodate asynchronous algorithms, support for communication between the Driver and Executors via RPCsthrough Maggy was added. The Optimizer that guides hyperparameter search is located on the Driver and it assignstrials to Executors. Executors periodically send back to the Driver the current performance of their trial, and theOptimizer can decide to early-stop any ongoing trial and send the Executor a new trial instead.

Contents: 1

maggy Documentation, Release 0.4.0

2 Contents:

CHAPTER 1

Quick Start

To Install:

>>> pip install maggy

The programming model consists of wrapping the code containing the model training inside a function. Inside thatwrapper function provide all imports and parts that make up your experiment.

There are three requirements for this wrapper function:

1. The function should take the hyperparameters as arguments, plus one additional parameter reporter which isneeded for reporting the current metric to the experiment driver.

2. The function should return the metric that you want to optimize for. This should coincide with the metric beingreported in the Keras callback (see next point).

3. In order to leverage on the early stopping capabilities of maggy, you need to make use of the maggy re-porter API. By including the reporter in your training loop, you are telling maggy which metric to reportback to the experiment driver for optimization and to check for global stopping. It is as easy as adding re-porter.broadcast(metric=YOUR_METRIC) for example at the end of your epoch or batch training step andadding a reporter argument to your function signature. If you are not writing your own training loop you canuse the pre-written Keras callbacks in the maggy.callbacks module.

Sample usage:

>>> # Define Searchspace>>> from maggy import Searchspace>>> # The searchspace can be instantiated with parameters>>> sp = Searchspace(kernel=('INTEGER', [2, 8]), pool=('INTEGER', [2, 8]))>>> # Or additional parameters can be added one by one>>> sp.add('dropout', ('DOUBLE', [0.01, 0.99]))

>>> # Define training wrapper function:>>> def mnist(kernel, pool, dropout, reporter):>>> # This is your training iteration loop>>> for i in range(number_iterations):

(continues on next page)

3

maggy Documentation, Release 0.4.0

(continued from previous page)

>>> ...>>> # add the maggy reporter to report the metric to be optimized>>> reporter.broadcast(metric=accuracy)>>> ...>>> # Return the same final metric>>> return accuracy

>>> # Launch maggy experiment>>> from maggy import experiment>>> result = experiment.lagom(map_fun=mnist,>>> searchspace=sp,>>> optimizer='randomsearch',>>> direction='max',>>> num_trials=15,>>> name='MNIST'>>> )

lagom is a Swedish word meaning “just the right amount”. This is how maggy uses your resources.

4 Chapter 1. Quick Start

CHAPTER 2

MNIST Example

For a full MNIST example with random search using Keras, see the Jupyter Notebook in the examples folder.

5

maggy Documentation, Release 0.4.0

6 Chapter 2. MNIST Example

CHAPTER 3

Documentation

Read our blog post for more details.

API documentation is available here.

3.1 Maggy User API

3.1.1 maggy.experiment module

Experiment module used for running asynchronous optimization tasks.

The programming model is that you wrap the code containing the model training inside a wrapper function. Insidethat wrapper function provide all imports and parts that make up your experiment, see examples below. Whenevera function to run an experiment is invoked it is also registered in the Experiments service along with the providedinformation.

maggy.experiment.lagom(map_fun, name=’no-name’, experiment_type=’optimization’,searchspace=None, optimizer=None, direction=’max’, num_trials=1,ablation_study=None, ablator=None, optimization_key=’metric’,hb_interval=1, es_policy=’median’, es_interval=300, es_min=10, de-scription=”)

Launches a maggy experiment, which depending on experiment_type can either be a hyperparameter optimiza-tion or an ablation study experiment. Given a search space, objective and a model training procedure map_fun(black-box function), an experiment is the whole process of finding the best hyperparameter combination inthe search space, optimizing the black-box function. Currently maggy supports random search and a medianstopping rule.

lagom is a Swedish word meaning “just the right amount”.

Parameters

• map_fun (function) – User defined experiment containing the model training.

• name (str) – A user defined experiment identifier.

7

maggy Documentation, Release 0.4.0

• experiment_type (str) – Type of Maggy experiment, either ‘optimization’ (default)or ‘ablation’.

• searchspace (Searchspace) – A maggy Searchspace object from which samples aredrawn.

• optimizer (str, AbstractOptimizer) – The optimizer is the part generating newtrials.

• direction (str) – If set to ‘max’ the highest value returned will correspond to the bestsolution, if set to ‘min’ the opposite is true.

• num_trials (int) – the number of trials to evaluate given the search space, each con-taining a different hyperparameter combination

• ablation_study (AblationStudy) – Ablation study object. Can be None for opti-mization experiment type.

• ablator (str, AbstractAblator) – Ablator to use for experiment type ‘ablation’.

• optimization_key (str, optional) – Name of the metric to be optimized

• hb_interval (int, optional) – The heartbeat interval in seconds from trial execu-tor to experiment driver, defaults to 1

• es_policy (str, optional) – The earlystopping policy, defaults to ‘median’

• es_interval (int, optional) – Frequency interval in seconds to check currentlyrunning trials for early stopping, defaults to 300

• es_min (int, optional) – Minimum number of trials finalized before checking forearly stopping, defaults to 10

• description (str, optional) – A longer description of the experiment.

Raises RuntimeError – An experiment is currently running.

Returns A dictionary indicating the best trial and best hyperparameter combination with it’s perfor-mance metric

Return type dict

3.1.2 maggy.searchspace module

class maggy.Searchspace(**kwargs)Create an instance of Searchspace from keyword arguments.

A searchspace is essentially a set of key value pairs, defining the hyperparameters with a name, type and afeasible interval. The keyword arguments specify name-values pairs for the hyperparameters, where values aretuples of the form (type, list). Type is a string with one of the following values:

• DOUBLE

• INTEGER

• DISCRETE

• CATEGORICAL

And the list in the tuple specifies either two values only, the start and end point of of the feasible interval forDOUBLE and INTEGER, or the discrete possible values for the types DISCRETE and CATEGORICAL.

Sample usage:

8 Chapter 3. Documentation

maggy Documentation, Release 0.4.0

>>> # Define Searchspace>>> from maggy import Searchspace>>> # The searchspace can be instantiated with parameters>>> sp = Searchspace(kernel=('INTEGER', [2, 8]), pool=('INTEGER', [2, 8]))>>> # Or additional parameters can be added one by one>>> sp.add('dropout', ('DOUBLE', [0.01, 0.99]))

The Searchspace object can also be initialized from a python dictionary:

>>> sp_dict = sp.to_dict()>>> sp_new = Searchspace(**sp_dict)

The parameter names are added as attributes of Searchspace object, so they can be accessed directly with thedot notation searchspace._name_.

add(name, value)Adds {name, value} pair to hyperparameters.

Parameters

• name (str) – Name of the hyperparameter

• value (tuple) – A tuple of the parameter type and its feasible region

Raises

• ValueError – Hyperparameter name is reserved

• ValueError – Hyperparameter feasible region in wrong format

get(name, default=None)Returns the value of name if it exists, else default.

get_random_parameter_values(num)Generate random parameter dictionaries, e.g. to be used for initializing an optimizer.

Parameters num (int) – number of random parameter dictionaries to be generated.

Raises ValueError – num is not an int.

Returns a list containing parameter dictionaries

Return type list

items()Returns a sorted iterable over all hyperparameters in the searchspace.

Allows to iterate over the hyperparameters in a searchspace. The parameters are sorted in the order ofwhich they were added to the searchspace by the user.

Returns an iterable of the searchspace

Type Searchspace

keys()Returns a sorted iterable list over the names of hyperparameters in the searchspace.

Returns names of hyperparameters as a list of strings

Type list

names()Returns the dictionary with the names and types of all hyperparameters.

Returns Dictionary of hyperparameter names, with types as value

3.1. Maggy User API 9

maggy Documentation, Release 0.4.0

Return type dict

to_dict()Return the hyperparameters as a Python dictionary.

Returns A dictionary with hyperparameter names as keys. The values are the hyperparametervalues.

Return type dict

values()Returns a sorted iterable list over the types and feasible intervals of hyperparameters in the searchspace.

Returns types and feasible interval of hyperparameters as tuple

Type tuple

3.1.3 maggy.callbacks module

class maggy.callbacks.KerasBatchEnd(reporter, metric=’loss’)A Keras callback reporting a specified metric at the end of the batch to the maggy experiment driver.

loss is always available as a metric, and optionally acc (if accuracy monitoring is enabled, that is, accuracy isadded to keras model metrics). Validation metrics are not available for the BatchEnd callback. Validation afterevery batch would be too expensive. Default is training loss (loss).

Example usage:

>>> from maggy.callbacks import KerasBatchEnd>>> callbacks = [KerasBatchEnd(reporter, metric='acc')]

class maggy.callbacks.KerasEpochEnd(reporter, metric=’val_loss’)A Keras callback reporting a specified metric at the end of an epoch to the maggy experiment driver.

val_loss is always available as a metric, and optionally val_acc (if accuracy monitoring is enabled, that is,accuracy is added to keras model metrics). Training metrics are available under the names loss and acc. Defaultis validation loss (val_loss).

Example usage:

>>> from maggy.callbacks import KerasBatchEnd>>> callbacks = [KerasBatchEnd(reporter, metric='val_acc')]

3.1.4 maggy.ablation module

class maggy.ablation.AblationStudy(training_dataset_name, training_dataset_version, la-bel_name, **kwargs)

The AblationStudy object is the entry point to define an ablation study with maggy. This object can subsequentlybe passed as an argument when the experiment is launched with experiment.lagom().

Sample usage:

>>> from maggy.ablation import AblationStudy>>> ablation_study = AblationStudy('titanic_train_dataset',>>> label_name='survived')

The above code will create an AblationStudy instance with a default dataset generator function, which uses theproject feature store to return a TFRecordDataset based on the feature ablation configuration (for an example,look at ablator.LOCO.get_dataset_generator()). If you want to provide your own dataset generator function,

10 Chapter 3. Documentation

maggy Documentation, Release 0.4.0

define it before creating the AblationStudy instance and pass it to the initializer. In the example below weassume the user has created a function called create_tf_dataset() that returns a TFRecordDataset:

>>> ablation_study = AblationStudy('titanic_train_dataset',label_name='survived', dataset_generator=create_tf_dataset)

In case you want to perform feature ablation with your custom dataset generator function, then of course yourfunction should be able to return specific datasets based on the feature ablation configuration. For an exampleimplementation of such logic, look at ablator.LOCO.get_dataset_generator().

After creating your AblationStudy instance, you should define your study configuration by including layers andfeatures that you want to be ablated:

>>> ablation_study.features.include('pclass', 'fare')>>> ablation_study.model.layers.include('my_dense_two',>>> 'my_dense_three')

You can also add a layer group using a list:

>>> ablation_study.model.layers.include_groups(['my_dense_two',>>> 'my_dense_four'])

Or add a layer group using a prefix:

>>> ablation_study.model.layers.include_groups(prefix='my_dense')

Next you should define a base model function using the layer and feature names you previously specified:

>>> # you only need to add the `name` parameter to layer initializers>>> def base_model_generator():>>> model = tf.keras.Sequential()>>> model.add(tf.keras.layers.Dense(64, activation='relu'))>>> model.add(tf.keras.layers.Dense(..., name='my_dense_two', ...)>>> model.add(tf.keras.layers.Dense(32, activation='relu'))>>> model.add(tf.keras.layers.Dense(..., name='my_dense_sigmoid', ...)>>> # output layer>>> model.add(tf.keras.layers.Dense(1, activation='linear'))>>> return model

Make sure to include the generator function in the study:

>>> ablation_study.model.set_base_model_generator(base_model_generator)

Last but not least you can define your actual training function:

>>> from maggy import experiment

>>> def training_function(dataset_function, model_function):>>> import tensorflow as tf>>> epochs = 5>>> batch_size = 10>>> tf_dataset = dataset_function(epochs, batch_size)>>> model = model_function()>>> model.compile(optimizer=tf.train.AdamOptimizer(0.001),>>> loss='binary_crossentropy',>>> metrics=['accuracy'])

3.1. Maggy User API 11

maggy Documentation, Release 0.4.0

>>> history = model.fit(tf_dataset, epochs=epochs, steps_per_epoch=30)>>> return float(history.history['acc'][-1])

Lagom the experiment:

>>> result = experiment.lagom(map_fun=training_function,>>> experiment_type='ablation',>>> ablation_study=ablation_study,>>> ablator='loco',>>> name='Titanic-LOCO')

__init__(training_dataset_name, training_dataset_version, label_name, **kwargs)Initializes the ablation study.

Parameters

• training_dataset_name (str) – Name of the training dataset in the featurestore.

• training_dataset_version (int) – Version of the training dataset to be used.

• label_name (str) – Name of the target prediction label.

to_dict()Returns the ablation study configuration as a Python dictionary.

Returns A dictionary with ablation study configuration parameters as keys (i.e. ‘train-ing_dataset_name’, ‘included_features’, etc.)

Return type dict

3.2 Maggy Developer API

As a developer you have the possibility to implement your custom optimizers or ablators. For that you can implementan abstract method, which you can then pass as an argument when launching the experiment. For examples, pleaselook at existing optimizers and ablators.

3.2.1 maggy.optimizer module

3.2.2 maggy.ablation.ablator module

class maggy.ablation.ablator.abstractablator.AbstractAblator(ablation_study, fi-nal_store)

finalize_experiment(trials)This method will be called before finishing the experiment. Developers can implement this method e.g.for cleanup or extra logging.

get_dataset_generator(ablated_feature, dataset_type=’tfrecord’)Create and return a dataset generator function based on the ablation policy to be used in a trial. Thereturned function will be executed on the executor per each trial.

Parameters

• ablated_feature (str) – the name of the feature to be excluded from the trainingdataset. Must match a feature name in the corresponding feature group in the feature store.

• dataset_type – type of the dataset. For now, we only support ‘tfrecord’.

12 Chapter 3. Documentation

maggy Documentation, Release 0.4.0

Returns A function that generates a TFRecordDataset

Return type function

get_number_of_trials()If applicable, calculate and return the total number of trials of the ablation experiment. Make sure to alsoinclude the base (reference) trial in the count.

Returns total number of trials of the ablation study experiment

Return type int

get_trial(ablation_trial=None)Return a Trial to be assigned to an executor, or None if there are no trials remaining in the experiment. Thetrial should contain a dataset generator and a model generator. Depending on the ablator policy, the trialscould come from a list (buffer) of pre-made trials, or generated on the fly.

Return type Trial or None

initialize()Initialize the ablation study experiment by generating a number of trials. Depending on the ablation policy,this method might generate all the trials (e.g. as in LOCO), or generate a number of trials to warm-startthe experiment. The trials should be added to trial_buffer in form of Trial objects.

3.3 Release 0.1

3.4 License

GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007. See LICENSE.

3.3. Release 0.1 13

maggy Documentation, Release 0.4.0

14 Chapter 3. Documentation

CHAPTER 4

Indices and tables

• genindex

• modindex

• search

15

maggy Documentation, Release 0.4.0

16 Chapter 4. Indices and tables

Python Module Index

mmaggy.experiment, 7

17

maggy Documentation, Release 0.4.0

18 Python Module Index

Index

Symbols__init__() (maggy.ablation.AblationStudy method),

12

AAblationStudy (class in maggy.ablation), 10AbstractAblator (class in

maggy.ablation.ablator.abstractablator),12

add() (maggy.Searchspace method), 9

Ffinalize_experiment()

(maggy.ablation.ablator.abstractablator.AbstractAblatormethod), 12

Gget() (maggy.Searchspace method), 9get_dataset_generator()

(maggy.ablation.ablator.abstractablator.AbstractAblatormethod), 12

get_number_of_trials()(maggy.ablation.ablator.abstractablator.AbstractAblatormethod), 13

get_random_parameter_values()(maggy.Searchspace method), 9

get_trial() (maggy.ablation.ablator.abstractablator.AbstractAblatormethod), 13

Iinitialize() (maggy.ablation.ablator.abstractablator.AbstractAblator

method), 13items() (maggy.Searchspace method), 9

KKerasBatchEnd (class in maggy.callbacks), 10KerasEpochEnd (class in maggy.callbacks), 10keys() (maggy.Searchspace method), 9

Llagom() (in module maggy.experiment), 7

Mmaggy.experiment (module), 7

Nnames() (maggy.Searchspace method), 9

SSearchspace (class in maggy), 8

Tto_dict() (maggy.ablation.AblationStudy method), 12to_dict() (maggy.Searchspace method), 10

Vvalues() (maggy.Searchspace method), 10

19